Measuring dietary diversity in resource poor countries: A decade of experience & remaining challenges

Transcription

1 Measuring dietary diversity in resource poor countries: A decade of experience & remaining challenges Mary Arimond Program in Community and International Nutrition University of California, Davis Megan Deitchler, Marie Ruel, Yves Martin Prevel ICDAM8, May 16, 2012

2 Why measure dietary diversity? Dietary diversity (DD) is one important dimension of dietary quality (also balance, moderation, etc.) All food based dnational dietary guidelines include this dimension, usually via recommended food groups for daily consumption This is based in part on idea that DD relates to micronutrient intakes/adequacy, which was well established established in developed countries (e.g. Kant, 1996 review)

3 Why measure dietary diversity(2)? Until very recently, lack of indicators has constrained programmatic action to improve dietary quality for vulnerable groups For infants, previous to 2008, no global indicators for quality of complementary diet during critical window 6 24 mo For women, repeated calls to action on women s nutrition but few programs beyond antenatal supplements, no indicators or yardsticks for progress in improving diet quality

4 Why measure dietary diversity(3)? For global tracking we need simplest possible indicators that meaningfully reflect diet quality Quantitative dietary data collection (e.g. 24 hr recalls) on nationally representative samples not affordable/feasible, esp. repeated across time Dietary diversity is feasible in large scale survey programs such as the DHS

5 Operationalizing dietary diversity DD has been defined & operationalized many ways Number of individual foods Number of food groups (with different levels of aggregation) Varying recall periods (yesterday & last week are most common) Indicators can be dichotomous ( % with low diversity ) or continuous scores

6 Measuring dietary diversity outline Recent progress impact of DD research Uptake of dietary diversity it indicators, to measure diet quality Dietary diversity, nutrient density & adequacy Do simple DD indicators predict nutrient density, adequacy? Remaining challenges Measurement issues Responsiveness Food groups & cut offs Consensus on uses

7 Global DD indicators: Two large collaborative projects Previously, a small handful of studies in resource poor settings* but no consistency in definitions, indicators Two large projects Working Group on Infant and Young Child Feeding Indicators Women s Dietary Diversity Project (WDDP) Both involved teams of researchers from many countries/institutions, and multiple data sets Assessed and compared multiple DD indicators using same definitions, analysis protocols across sites * Review Ruel, 2003; Hatloy et al. 1998; Torheim et al. 2004; after initiation of IFPRI projects: Steyn et al. 2006; Kennedy et al. 2007; Daniels et al. 2009

8 WHO IYCF indicators for global luse US Government Feed the Future Initiative Both include infant dietary diversity indicator; FTF also includes women s DD

9 FAO revised guidelines, 2011 Guidance on measuring Guidance on measuring individual DD harmonized with one of the best performing WDDP indicators

10 IYCF DD indicator for 6 23 mo Percent with 4 or more food groups WHO, 2008 Ethiopia Niger Burkina Faso Mali DRC Mozambique Kenya Ghana Bangladesh India Cambodia Indonesia Haiti Colombia Bolivia Peru

11 Dietary diversity and nutrient density (Working Group on IYC Feeding Indicators 2006) Breastfed infants 6 8 mo, MMDA by # food groups yesterday MMDA MMDA is a measure of the adequacy of nutrient density, relative to needs, and averaged across 9 problem nutrients

12 Dietary diversity & micronutrient adequacy (Arimond et al., Women s Dietary Diversity Project, J Nutr. 2010) MPA is probability of adequacy averages across 11 micronutrients

13 Classifying individuals vs. assessing population prevalence Infants: we assessed cut offs to predict MMDA above 75%. For selected cut off of 4 groups out of 7, misclassificationranged from 5 45% (across 10sites and 3 age sub groups) Women: We assessed cut offs to predict MPA above 70%. No cut off was selected; for 5 or more groups out of 9, misclassification ranged from 14 42% At population level, best choice indicator predicted prevalence reasonably well (within +/ 6%)

14 Summary of relative lti validity Across allstudies and all DD indicators, there are moderate to strong associations with nutrient density and/or nutrient intakes: DD indicators are robust Indicator performance (sensitivity, specificity, and best cut offs) vary by context and all indicators result in some misclassification: DD indicators are imperfect This has implications for uses

15 Challenges measurement issues One WDDP study compared DD indicators derived from simple qualitative recall to more quantitative approach (24 hr recall) Showed substantial differences between the two Problems increased with more disaggregated indicators (e.g., 13 and 21 group indicators) Questionnaires must be carefully designed and tested to excludefoods consumed in trivial amounts Additional methodological work is warranted (Martin-Prevel, Bequey, Arimond. J Nutr. 2010)

16 Challenges responsiveness DD indicators have been proposed for use in assessment and monitoring progress at national and sub national levels Yet little is known about responsiveness of the indicators to interventions, shocks Study of 2008 food price shock in Burkina Faso showed impact on DD (next slide) Inclusion of indicators at baseline andpost Inclusion of indicators at baseline and postintervention in Feed the Future projects will yield more insights (18 countries, large geographic areas)

17 Response to spike in food prices* Ouagadougou Nb of food groups Decrease in dietary diversity P< Mean DDS 2008 (Adjusted) Food groups that showed the largest percent decrease were nutrient-dense: Dairy, meat, poultry, fruits, vitamin A-rich vegetables, nuts & seeds *(Martin-Prevel et al, presented FENS Madrid 2011)

18 Challenges consensus on food groupings & on cut offs Dichotomous indicators are likely necessary for advocacy and communication with policy makers Decisions on specific indicator definitions are informed by study results, but also by considerations of feasibility and communication For IYCF indicators, WHO facilitated a multi, stakeholder process consensus on indicator definitions and cut offs (including DD indicator)

19 Challenges consensus on uses Simple DD indicators (across all studies) show only moderate sensitivity andspecificity for classifying intakes; never recommended for individual level use (e.g. never recommended for screening) Recommended for use at population level, but due to measurement error only when sample sizes are reasonably large For repeat uses in same population, and for p p p, comparisons between populations, attend to seasonality

20 Conclusions DD indicators are imperfect but imperfect indicators can bevery useful (e.g.: exclusive breastfeeding indicator) DDindicators arerobustrobust many different versions of indicators consistently associated with MN adequacy Going forward need to bl balance indicator improvement (e.g. improvements in measurement) against the value of having comparable measurement over time in data series